Template

%load_ext pretty_jupyter

Conda¶

.tabset|O_O|.tabset-pills|O_O|.tabset-fade

miniconda is a light version of anaconda, it contains only conda, python and other necessary tools. miniforge is similar to minicoda but conda-forge is the only channel by default. miniconda has package installer while miniforge has only command installer.

Use conda-forge channel instead of defaults channel when installing packages and use pip only when conda doesn't have certain packages.

记住一句话：类是模板，而实例则是根据类创建的对象。我初学时对类的理解是从类的字面上，可以片面的认为它是一个种类，它是相似特征的抽像，也就是相似的东西，可以把相似特征的事务抽象成一个类。（事务可以是具体的物体或行为）以圆为例，圆是具有圆周率(pi)和半径(r)两个相似特征的属性。根据相似特征抽象出圆类，每个圆的半径可以不同，那么半径可以作为圆的实例属性；而每个圆的圆周率pi是相同的，那么圆周率pi就可以作为类属性，这样就定义出了一个圆类。而我们要知道圆的面积，周长等可以通过类方法计算出来。

作者：木头人 链接：https://zhuanlan.zhihu.com/p/30024792 来源：知乎 著作权归作者所有。商业转载请联系作者获得授权，非商业转载请注明出处。

Install conda¶

First, go to miniconda installer, download macos.pkg and install it.

Second, check if conda is configured using conda --version. If not, it needs to be configured in file .zshrc (a place for personal configurations):

• Open this file in terminal using vim ~/.zshrc,
• Type i to open edit mode (--INSERT-- will appear at bottom),
• Write export PATH=/Users/chengfengliu/Softwares/miniconda3/bin:$PATH in a new line,
• Press esc to close edit mode,
• Move the cursor to end and type : (: will appear at bottom), type wq (add ! if there is readonly problem) after : and press enter to close .zshrc,
• Make the edited configurations effective using source ~/.zshrc,
• Check again using conda --version.

Third, set conda-forge as the only channel:

• Check current channels with priorities using conda config --get channels,
• Add conda-forge channel with the highest priority using conda config --add channels conda-forge,
• Set channel priority strict using conda config --set channel_priority strict,
• Remove defaults channel using conda config --remove channels defaults,
• Check again using conda config --get channels,
• In base env, update all packages from defaults channel to conda-forge channel using conda update --all,
• Check by conda list.

If xcrun error occurs, use xcode-select --install.

Conda commands¶

conda --version
conda list

conda search --full-name pkg_precise
conda search pkg_pattern

conda install pkg=version
conda uninstall pkg

conda update conda
conda update pkg=version
conda update --all

conda create --name env python=version
conda create --name env1 --clone env2
conda activate env
conda deactivate
conda env list
conda env remove --name env

conda config --get channels
conda config --add channels channel
conda config --set channel_priority strict
conda config --remove channels channel

Pip¶

.tabset|O_O|.tabset-pills|O_O|.tabset-fade

Install pip¶

Installing python using conda when creating an env comes with pip.

Mac uses zsh in terminal which causes no matches found issue with spuare brackets, solution follows:

• Open .zshrc file in terminal using vim ~/.zshrc,
• Type i to open edit mode (--INSERT-- will appear at bottom),
• Write setopt no_nomatch in a new line,
• Press esc to close edit mode,
• Move the cursor to end and type : (: will appear at bottom), type wq (add ! if there is readonly problem) after : and press enter to close .zshrc,
• Make the edited configurations effective using source ~/.zshrc.

Pip commands¶

pip --version
pip list
pip list --outdated

pip check pkg
pip check

pip install pkg==version
pip install --upgrade pip
pip install --upgrade pkg==version
pip uninstall pkg

pip freeze
pip freeze > requirements.txt
pip install -r requirements.txt

Environments¶

Using Jupyter extension in VSCode doesn't require Jupyter Lab or Notebook installed because they are user interfaces just like VSCode, but the following packages are needed:

• ipykernel converts conda environments into kernels
• nbconvert with pretty-jupyter exports Jupyter notebooks into htmls

Create data environment in base environment:

conda create --name data ipykernel \
    python python-docx pandas openpyxl \
    scikit-learn py-xgboost lightgbm catboost optuna \
    langchain openai transformers \
    plotly nbconvert

pip install pretty-jupyter

To activate data environment and enter into Projects directory when launching terminal:

• Open .zshrc file in terminal using vim ~/.zshrc,
• Type i to open edit mode (--INSERT-- will appear at bottom),
• Write conda activate data and cd /Users/chengfengliu/Documents/Projects in separated new lines,
• Press esc to close edit mode,
• Move the cursor to end and type : (: will appear at bottom), type wq (add ! if there is readonly problem) after : and press enter to close .zshrc,
• Make the edited configurations effective using source ~/.zshrc.

Plotly¶

import plotly.express as px

df = px.data.iris()

fig = px.scatter_matrix(df, 
                        dimensions=["sepal_width", 
                                    "sepal_length", 
                                    "petal_width", 
                                    "petal_length"], 
                        color="species")

fig.show(renderer='notebook')